Abstract
This paper proposes a multi-path effect analysis model for major economies in the global trade network based on the MH-DP Transformer architecture. The global trade network comprises interconnected nodes, with interactions among major economies (China, the US, Japan, and Europe) being crucial. A trade network model with four regional nodes and their connections is established to simulate trade flow propagation and capture the nonlinear feedback mechanism through an iterative process. The experimental results demonstrate that trade volume changes exhibit significant nonlinear trends, and mutual influences between economies produce strong cascading effects. For instance, China experiences significant fluctuations when interacting with the US and Europe, while Europe shows strong chain reactions with Japan. The model outperforms other comparative models, achieving the best path prediction accuracy (82.5%) and convergence speed, effectively capturing nonlinear effects. Furthermore, the paper uses 3D charts and heat maps to visualize global trade flow changes, offering a quantitative perspective on understanding the complexity of global trade interactions. The experimental results validate the model's strong ability to analyze multi-path effects and dynamic chain reactions in global trade networks, providing a powerful tool for future economic policy formulation and interactive prediction in regional economic integration.